Correlated Differential Privacy Protection for Big Data
Denglong Lv, Shibing Zhu · 2018
Aiming at the privacy leakage problem of traditional differential privacy function in correlated datasets, a novel improved method based on machine learning and maximum information coefficient (MIC) was proposed, which improved the difference privacy of correlated datasets in big data. On this basis, the r-CBDP (r-Correlated Block Differential Privacy) protection model was proposed. Firstly r-CBDP used MIC and machine learning to determine the dependencies between correlated datasets, accurately calculated the correlated sensitivity of the query function, and then used clustering to divide big data into independent blocks, and implemented r-CBDP protection of data blocks to achieve the whole big data correlated differentiated privacy protection.